PhD Candidate · WPI · Boston, MA

Steven
Hyland.

I build robots that reason about the physical world — specializing in manipulation, dynamics modeling, and bridging simulation to reality. Currently finishing a PhD in Robotics Engineering at WPI.

Steven Hyland
3 Publications
4+ Years Research
Robots Deployed

Physics-first.
Learning-forward.

I'm a PhD candidate at Worcester Polytechnic Institute, where I research active perception and parameter estimation for robot manipulation. My work focuses on letting robots figure out the physical properties of objects they've never encountered before — without being told in advance.

My background is rooted in analytical, model-based methods — robot dynamics, sim2real transfer, and model-based control. I'm increasingly combining these with modern machine learning to build systems that are both physically grounded and data-driven.

Before the PhD, I deployed and programmed autonomous mobile robots for Fortune 500 clients at Seegrid. I hold a B.S. in Mechanical Engineering from Columbia University. I'm looking for research engineer roles where I can build systems that actually work — in the real world, not just on paper.

Robot Dynamics Manipulation Sim2Real Parameter Estimation MuJoCo ROS 1 & 2 Python / PyTorch C++ MATLAB
Presidential Fellow
Worcester Polytechnic Institute
NSF NRT FORW-RD Fellow
National Science Foundation
Glenn Yee Award Recipient
Worcester Polytechnic Institute
IEEE RAS President
WPI Student Branch
Kings Crown Leadership Award
Columbia University
Research

Current work.

IEEE CASE 2025 · Primary
CoM Estimation with Onboard Sensing & Pushing for Mobile Robots
A framework for estimating the 3D center of mass of unknown payloads using a holonomic mobile robot. Combines active pushing, onboard sensing, and model-based estimation — no prior knowledge of the object required.
IEEE IROS 2023
Predicting Center of Mass by Iterative Pushing
Iterative pushing strategy to infer object CoM for transportation and manipulation tasks. Validated on physical hardware with a robot arm, demonstrating robust estimation across payload geometries.
Ongoing · WPI
Learning-Based Dynamics for Manipulation Planning
Extending model-based CoM estimation with reinforcement learning and learned dynamics in MuJoCo. Goal: generalize across payload shapes and surface properties without hand-crafted physical models.
Projects

Built over time.

2025
Online Parameter Estimation with RL (In Progress)
MuJoCo RL Python Stable Baselines3

Simulation environment in MuJoCo where a robot arm must estimate the mass and inertia of unknown objects during a manipulation task — using reinforcement learning rather than analytical estimators. Extension of PhD work into learned dynamics.

2024
Graph-Based Optimal Grasp Configuration
Python Graph Search MoveIt ROS 2

Designed and validated a graph-based grasp planner using a modified breadth-first search algorithm. Selects optimal arm configurations for payload manipulation based on estimated CoM, minimizing torque and improving stability.

2023
OddRugs — Optimal Dynamic Distribution of Robots via Graph Search
Python Multi-Robot Graph Search

Multi-robot task allocation system using graph-based search to dynamically distribute robots across a workspace. Optimizes coverage and throughput under real-time constraints.

2023
Constraint-Based 3D Haptic Interaction Simulation
C++ Haptics Simulation

Implemented a constraint-based haptic simulation framework for 3D interaction. Models contact forces and physical constraints to produce realistic tactile feedback in a virtual environment.

2020 – 2021
AMR Fleet Deployment Tools — Seegrid
Python Fleet Management AMR

Built internal Python tooling at Seegrid to automate and accelerate AMR deployment for Fortune 500 clients. Reduced per-installation time by ~48 hours. Also designed fleet pathing logic for 20+ robot sites.

2019
Tensegrity Tumbleweed Rover — Columbia Capstone
Mechanical Design SolidWorks Prototyping

Senior capstone project at Columbia: designed and built a tensegrity-based rolling robot for planetary exploration. The structure uses tensile integrity to survive impacts without rigid housing.

Experience

Where I've worked.

Aug 2021
— Present
Doctoral Research Assistant
Worcester Polytechnic Institute
  • Developed 3D CoM estimation framework combining active perception, non-prehensile pushing, and model-based control for arbitrary payloads — published at IROS 2023 and CASE 2025.
  • Built high-fidelity simulation pipeline in MuJoCo; extending with RL-based learned dynamics for generalization.
  • Designed graph-based optimal grasp configuration planner using modified BFS.
  • Mentored undergraduate researchers; led outreach for NSF FORW-RD and IEEE RAS.
Nov 2019
— Aug 2021
Implementation Engineer
Seegrid · Pittsburgh, PA
  • Programmed and deployed AMRs for Fortune 500 clients including Amazon, GM, and UPS.
  • Designed fleet pathing logic for 20+ robot installations to maximize throughput.
  • Built Python tooling that cut per-site deployment time by ~48 hours.
Sep 2019
— Nov 2019
Robotic Design Engineer
Avar Robotics · New York, NY
  • Modeled next-generation inventory-sorting robot in SolidWorks with FEA validation.
  • Prototyped tetherless mobile redesign enabling infrastructure-free operation.
Publications

Peer-reviewed work.

2025
Onboard Sensing and Pushing of Unknown Payloads for CoM Estimation with a Holonomic Mobile Robot
IEEE CASE 2025 · Los Angeles, CA
PDF ↗
2025
Onboard Sensing and Pushing of Unknown Payloads for CoM Estimation with a Holonomic Mobile Robot
New England Manipulation Symposium · Boston, MA
PDF ↗
2023
Predicting Center of Mass by Iterative Pushing for Object Transportation and Manipulation
IEEE IROS 2023 · Detroit, MI
PDF ↗
Skills

What I work with.

Robotics
  • Robot Dynamics Modeling
  • Sim2Real Transfer
  • Manipulation Planning
  • Parameter Estimation
  • Model-Based Control
  • Active Perception
  • Fleet Management
Software & Tools
  • Python (NumPy, SciPy, PyTorch)
  • MuJoCo
  • ROS 1 & 2
  • C++
  • MATLAB
  • Gazebo
  • MoveIt
ML / AI
  • Reinforcement Learning
  • Imitation Learning
  • Learned Dynamics
  • Stable Baselines3
  • PyTorch
  • VLMs (exploring)
Engineering
  • Mechanical Design
  • SolidWorks & FEA
  • Prototyping
  • Scientific Communication
  • Cross-functional Collab
  • Mentorship & Leadership
Contact

Let's talk robots.

I'm actively looking for research engineer and applied scientist roles in robotics and robot learning, based in Boston. Open to relocation for the right opportunity.